The single biggest difference between people who find AI useful and people who find it disappointing isn't the tool. Everyone has roughly the same models. It is how they ask.
Most people type one line and hope. That one line is usually the instruction and nothing else, which is like handing a new hire a sticky note that says "make the newsletter" and then being surprised by what comes back.
I teach a framework for this called RIPEE. It is five things, and none of them are technical. If you have ever delegated well to a person, you already know all five.
R: Role
Tell it who to be.
"You are an event planner with ten years of experience running fundraisers for nonprofits."
This isn't a costume. It changes what the model reaches for. Ask a generic assistant about an event and you get generic event advice. Ask an experienced fundraiser and you get thinking about donor flow and run-of-show and the awkward gap between dinner and the ask.
I: Instructions
What you actually want. This is the only part most people include, which is why most results are mediocre.
Be more specific than feels necessary. "Write a newsletter" and "write a 400-word update for our volunteers about the schedule change, leading with the reason" produce very different things.
P: Parameters
The constraints. Length, format, tone, budget, audience, what to avoid.
This is where most of the quality lives, and it's the step people skip because it feels like extra typing. Those few extra seconds are the whole difference between output you can use and output you have to rewrite.
If you would tell a person "keep it under a page and do not make it sound like a corporate memo," tell the AI.
E: Examples
Show it what good looks like.
Paste a previous newsletter you liked. A competitor's page. Two sentences in the voice you want. It doesn't have to be polished. Rough is fine.
This is the highest-leverage step by a wide margin and the one almost nobody does. Models are extraordinarily good at pattern matching and comparatively bad at guessing your taste. One example beats three paragraphs of description.
E: Emotion
Why it matters.
"This is how new families decide whether to visit our church." "This is the first thing a grieving parent will read."
This one sounds soft and it isn't. Stakes change the weight of every word. It is also the step that makes you articulate, to yourself, what the thing is actually for. I've caught my own muddled thinking at this step more than once.
Putting it together
Take a real example. The cold version:
Make a website for my club.
Generic. Could belong to any club anywhere. No names, no times, no reason for anyone to show up.
The RIPEE version:
R: You are an award-winning web designer for student organizations. I: Build a one-page site for The Venture Club, the business club at Summit Ridge High. One HTML file, styling included. P: Hero and tagline, Who We Are, What We Do, meeting time Thursdays 3:30 in Room 214, and a Join button. Navy and gold. Mobile friendly. E: Should feel like a confident startup landing page, not a school bulletin board. E: This is how new students decide whether to show up.
Same model. Same thirty seconds of typing, maybe sixty. Completely different output, and the second one is usable.
Why the framework is really about you
Here's the thing I did not expect when I started teaching this.
RIPEE is a thinking tool disguised as a prompting tool.
You cannot fill in Role without deciding what kind of expertise the problem needs. You can't fill in Parameters without knowing your constraints. You cannot fill in Emotion without knowing who this is for and why it matters to them.
People arrive at the framework wanting better AI output. What they often get first is clarity about the task, and then the AI output improves as a side effect.
Which points at something worth holding onto. The temptation with these tools is to let them do the thinking. The way to use them well is almost the opposite: they reward you for having thought more clearly than you otherwise would have.
Bad input, bad output. Better prompt, better result. But "better prompt" mostly means "you understood your own request."
Try it on one thing this week. Take a request you would normally fire off in a single line, and give it all five. The difference isn't subtle.
RIPEE is the framework I teach in workshops and in my classroom. Use it, adapt it, name it whatever you like.